O1: insert attack points

one_run_audit/audit.py

@@ -205,15 +205,39 @@ def get_dataloaders_raw(m=1000, train_batch_size=512, test_batch_size=10):
     train_ds = CIFAR10(root=DATADIR, train=True, download=True)
     test_ds = CIFAR10(root=DATADIR, train=False, download=True)
 
-    train_x = preprocess_data(train_ds.data)
+    train_x = train_ds.data
-    test_x = preprocess_data(test_ds.data)
+    test_x = test_ds.data
-    train_y = torch.tensor(train_ds.targets)
+    train_y = np.array(train_ds.targets)
-    test_y = torch.tensor(test_ds.targets)
+    test_y = np.array(test_ds.targets)
+
+    mask = np.full(len(test_x), False)
+    mask[:m] = True
+    mask = mask[np.random.permutation(len(test_ds))]
+    S = np.random.choice([True, False], size=m)
+
+    attack_x = test_x[mask][S]
+    attack_y = test_y[mask][S]
+
+    for i in range(len(attack_y)):
+        while True:
+            c = np.random.choice(range(10))
+            if attack_y[i] != c:
+                attack_y[i] = c
+                break
+
+    train_x = np.concatenate([train_x, attack_x])
+    train_y = np.concatenate([train_y, attack_y])
+
+    train_x = preprocess_data(train_x)
+    test_x = preprocess_data(test_x)
+    train_y = torch.tensor(train_y)
+    test_y = torch.tensor(test_y)
 
     train_dl = DataLoader(
         TensorDataset(train_x, train_y.long()),
         batch_size=train_batch_size,
         shuffle=True,
+        drop_last=True,
         num_workers=4
     )
     test_dl = DataLoader(
@@ -494,7 +518,6 @@ def train_fast(hp):
     )
 
     train_no_cap(model, hp, train_dl, test_dl, optimizer, criterion, scheduler)
-
     return init_model, model
 
 def train(hp, train_dl, test_dl):